The primary goal of this project was to determine the extent to which the adoption and implementation of clinical practice guidelines lead to changes in the costs of providing care within a healthcare system.
Our project had two primary aims: (1) classify VA facilities according to the intensity of DM guideline implementation; and (2) determine the extent to which the intensity of DM guideline implementation at VA facilities was associated with differences in costs and utilization.
The empirical testing of these aims made use of a unique research tool-a comprehensive VA database created for the purposes of this study that linked together a large national cohort of well-characterized patients in the 1999 Large Health Survey of Veteran Enrollees (LVHS), national guideline, quality improvement and organizational data collected by our study group, and cost data developed by the VA Health Economics Resource Center (HERC). The 1999 LVHS served as the study cohort. We focused on patients with the primary diagnosis of DM, confirmed by rigorous algorithms that considered clinical and administrative data on utilization, testing and treatment. LVHS respondents were allocated to VAMCs in which they received most of their care within a given year, and were tracked over the six-year period 1999-2004. We developed a summative scoring scheme based on item response theory (IRT). We estimated fixed and random effects panel data models of (1) the probability of one or more DM-related inpatient admissions and (2) DM-related inpatient treatment costs. Two-stage least squares regression was used to address the potential endogeneity of the hospital-level IRT guideline measure.
Fixed-effects model showed that hospitals with sicker patients also tended to have higher (i.e., better) guideline adherence. Fixed-effects two-stage lest-squares model showed that lagged IRT was associated with significantly lower probability of DM inpatient admission. However, IRT had no statistically significant effects on DM inpatient cost. Higher comorbidity scores were associated with significantly higher inpatient admission probability and higher inpatient costs. The first-stage F values indicated that IVs highly correlated with IRT. Over-identifying restriction tests for both models suggested that IVs did not correlate with unmeasured confounders (p ≤ 0.05). In sensitivity analyses, two-stage least-squares random-effects models showed that lagged IRT was associated with significantly lower probability of DM admission and significant lower levels of DM costs. Random-effects IV cost models for IRT were similar to fixed-effects models.
The results will in turn enable clinicians, managers, and policy makers to gain a better understanding of the resource implications associated with changes in clinical and organizational structures and processes, and provide a business case for improvements in DM guideline adherence.
External Links for this Project
None at this time.